There is great scientific and societal interest in the ecology and functioning of the immense diversity of microorganisms associated with plant roots (Mendes et al., 2011; Porras-Alfaro & Bayman, 2011). In particular, research into plant–soil interactions has unveiled a pivotal role of root-associated fungi in influencing plant growth and community structure (van der Heijden et al., 2008; Schnitzer et al., 2011; Wagg et al., 2014). So far, knowledge on the identity of fungi associated with plant roots, and forces structuring the communities they form, is still scarce. This extends to agricultural systems, where communities of belowground fungi are a largely unknown but potentially important driver of plant productivity akin to natural systems, and display a considerably high diversity (Orgiazzi et al., 2012). So far, most research has focused on plant pathogens (e.g. Xu et al., 2012) and on arbuscular-mycorrhizal fungi (AMF). AMF are an important group of plant symbionts, and we know that these generally increase in diversity in response to reduced agricultural management intensity (Oehl et al., 2004; Verbruggen et al., 2012).
For other groups of root endophytes little is known about responses to agricultural management, even though they may be of high ecological significance (Rodriguez et al., 2009). Apart from potential effects on plants, there is great interest in identifying taxa that may serve as bio-markers for sustainable agricultural practices, as has recently been explored for AMF by Jansa et al. (2014). So far this has not been attempted for other root inhabiting fungi, likely because it is unknown whether root-colonizing fungi are sensitive to changes in land-use intensity. In this study we have sampled wheat roots in agricultural fields that were either managed conventionally (seven sites) or had been converted to organic farming recently (2–4 yr; eight sites), moderately long ago (10–14 yr; six sites), or had been subjected to long-term organic farming (16–33 yr; eight sites). We analyzed the fungal community in roots using next generation sequencing of fungi and ask how different biotic and abiotic aspects drive fungal communities inhabiting wheat roots.
Materials and Methods
All sites were situated in Switzerland occurring in a circle with a radius of c. 50 km around Zürich, northern Switzerland, with a similar soil type typical of the region (cambisol with a loamy texture; see Honegger et al., 2014, for further description). All sites contained pasture in 2010, grew maize (Zea mays) in 2011 and winter-wheat (Triticum aestivum) or spelt (Triticum aestivum ssp. spelta) from 2011 to 2012 (see Supporting Information Notes S1 for further details on sampling and molecular analysis). In short: roots from six wheat individuals were sampled between 4 and 8 June 2012, washed, DNA was isolated from 7 to 10 mg lyophilized roots, and subjected to PCR using the general fungal primers ITS1F (Gardes & Bruns, 1993) and ITS4 (White et al., 1990). Additionally, at each field 15 soil samples were taken using a corer (3 cm width, 20 cm depth), pooled, and used to estimate the following parameters: pH (H2O), available phosphorus (P; CO2 -extractable), available potassium (K; CO2 -extractable) and magnesium (Mg; CaCl2 -extractable). Between 26 and 29 June soil cover for each plant species (weeds) was determined in all fields (except for one organically managed field) in three randomly placed 1-m2 plots. Cover of each plant was averaged over these three plots, Shannon diversity (H) was calculated based on this average per-species cover and averages were summed as an estimate of total plant cover.
Read numbers of operational taxonomic units (OTUs; see Fig. S1 for sample-based rarefaction plots) were natural log (loge + 1) transformed for all analyses, and Bray Curtis dissimilarity index (abundance based) and Jaccard index (presence/absence based) were used to assess community differences between sites. As a measure of environmental distances between sites, Principal Component Analysis was performed on normalized values of the variables soil pH, available P, K and Mg, soil respiration and microbial biomass as representing the ‘soil environment’ (the latter two biotic variables are included because they may be inclusive of environmental differences not covered by the measured abiotic variables). The Euclidean distance between site projections along the first three Principal Components was calculated (weighted according to eigenvalues of respective axes) which together accounted for 92.1% of variance, and used as a measure of site environmental dissimilarity. As a measure of vegetation dissimilarity absolute cover (in m2) by each weed species was natural log (loge + 1) transformed and for the resulting matrix Euclidean distances between sites were calculated, because this measure explicitly puts weight on abundances (cover) of species. For spatial distance the Euclidean distance between each site was calculated in meters using the geographical coordinates determined at each site. Each of these dissimilarity matrices was compared to fungal community dissimilarity using a Mantel test with 9999 permutations.
Different management groups were compared using PERMANOVA with the four different times since conversion (0, 2–4, 10–14, > 16 yr) as factor levels. In order to identify which environmental and biotic variables significantly predict fungal community composition, we performed redundancy analysis (RDA, euclidean distance based) with a combination of forward and backward predictor selection using the ordistep function in R. This procedure tests which predictors significantly explain the community using a permutation test, takes the strongest one as a covariate, and repeats this process until no further predictor significantly improves the model. In between each addition of a covariate it additionally tests whether previous additions are still significant. The predictors tested in this model were: pH, available P, Mg, K, plant H, and plant cover. Soil respiration and microbial biomass were not included because they were strongly collinear with the abiotic variables; however, an additional test indicated that inclusion of these variables would not have affected the results. If not specifically mentioned otherwise analyses were performed in R using the vegan package 2.0-5 (Oksanen et al., 2012).
The different management groups were found to not have a significant effect on fungal communities (abundance: pseudo-F3,25 = 0.99; P =0.46, presence/absence: pseudo-F3,25 = 0.98; P =0.50), which means that community dissimilarity (turnover) was as large within groups as between groups. The three most common OTUs across fields were a taxon closely matching Glarea lozoyensis which occurred in all fields, a member of the Sordariomycetes also occurring in all fields, and another Sordariomycete related to the Lasiosphaeriaceae occurring in all but two fields. Relative abundances and occurrence frequencies of the most frequent OTUs are shown in Fig. 1(a). Relating fungal community dissimilarity to site-specific parameters showed a significant effect of soil environmental dissimilarity (R =0.25, P =0.017), but not of spatial distance or vegetation composition, and this was the same using a presence/absence based distance measure (environment: R =0.24, P =0.019; space and vegetation: not significant (NS)). RDA testing the effect of the separate environmental and biotic predictors on OTUs resulted in an explained variance of all predictors of 33.1%, where soil pH (F =3.45, P <0.01), Mg concentration (F =2.04, P <0.01) and plant H (F =1.41, P =0.017) were sequentially selected as significant predictors. Together these predictors explained 22.9% of variance, and no single further predictor addition improved the model. When based on presence/absence of fungal taxa, pH and Mg were also sequentially selected but plant diversity ceased to significantly improve the model (results not shown).
Interestingly we found seven representative OTUs of Sebacinales occurring in a total of 11 sites, exclusively on organic farms. Relative abundances of these OTUs were low, ranging from 0.01% to 0.44% of the number of reads per site. In order to test whether management intensity may affect occurrence of Sebacinales, we performed a nested PCR with the same DNA samples using the Sebacinales-specific primers NSSeb1 – NL2R and NSSeb2 – NLSeb1.5R (Garnica et al., 2013). We screened PCR products of this reaction on a 1.5% agarose gel for presence of bands, cleaned the products and subjected the positive ones to Sanger sequencing using the NSSeb2 primer. We tested the effect of management (organic vs conventional) using a χ2 test (P-value was approximated using Monte Carlo permutation because of low values under conventional management).
We yielded amplification product in all but one of the 11 organic sites where Sebacinales were detected previously, but also in six more sites bringing the total of sites to 17. As can be seen in Fig. 1(b), these were almost exclusively (except for one) organically managed fields. Sanger sequencing of these 16 samples yielded 12 good quality sequences, all of which were confirmed to be of Sebacinalean origin by BLAST analysis. These sequences represented eight unique OTUs (97% similarity), of which one was shared by four sites, another by two, and the other six only occurred in one site each (see Table S1 for BLAST scores of sequences). The remaining four samples are inferred to contain multiple Sebacinales sequences and could therefore not be identified by direct Sanger sequencing. Phylogenetic analysis with representative sequences obtained from Weiss et al. (2011) and Garnica et al. (2013) showed that all OTUs clustered with Sebacinales group B (not shown). Logistic regression (GLM function with binomial error distribution based on presence/absence) indicated that occurrence of Sebacinales across sites is significantly predicted by organic vs conventional agriculture (z = 2.35, P =0.019) and by plant H, but not by other environmental predictors when used as predictors separately (Table 1). When both predictors are added simultaneously, plant diversity still remains a significant predictor (z = 2.01, P <0.05).
Table 1. Relationship between the occurrence of Sebacinales members and field specific environmental predictors using logistic regression
We found that total fungal communities in roots of wheat plants were primarily driven by soil abiotic predictors, in particular soil pH and Mg concentration. This is in line with other studies on fungal communities in plant roots and soil (e.g. Widden, 1987; Dumbrell et al., 2010; Rousk et al., 2010), where pH has been found to be a main driver. In case of Mg concentration, coarse-scale fungal community shifts in arbuscular mycorrhizal fungi and dark septate endophytes (DSE) have been reported (Postma et al., 2007). In addition to these predictors, we found that the diversity of weed communities significantly predicted the fungal community inside wheat roots. Earlier work investigating AMF communities in roots showed that the neighboring plant community can have a strong effect (Mummey et al., 2005; Hausmann & Hawkes, 2009). In the current study, AMF were detected at a relatively low read abundance (0.12%) precluding specific analysis, which may be caused by a bias of the ITS1F-ITS4 primer set in favor of Dikarya (e.g. Orgiazzi et al., 2012). For taxa other than AMF and ectomycorrhizal fungi (Simard et al., 2012), the influence of surrounding plant communities on root-fungal communities is much less known and may very well often be absent (Botnen et al., 2014). The fact that the Sebacinales (Weiss et al., 2011) and other root associated fungi have a broad host range can cause the surrounding plant community to influence fungal communities in wheat roots, if these fungi proliferate more extensively in some hosts other than wheat. Such a mechanism could result in weed diversity significantly changing fungal communities in wheat roots, and potentially explain the strong relationship between Sebacinales occurrence and weed diversity we report here (Table 1). An alternative reason for the effect of weeds on fungi found in wheat roots is that weed litter left in soil stimulates growth of fungi that can colonize wheat roots, as many root endophytes are known to have saprobic activity (Porras-Alfaro & Bayman, 2011). This is also true for Sebacinales (Zuccaro et al., 2011), of which some members form ericoid mycorrhizas and thus clearly associate with soil very rich in organic matter. Therefore, more work is needed to establish whether plant diversity is indeed responsible for significant changes in root-colonizing fungal communities or whether there may be other underlying factors responsible for these relationships.
One of the main findings we report here is that fungi in the order of Sebacinales were significantly more prevalent in wheat roots from organically managed than from conventionally managed fields. As can be seen in Fig. 1(b), their occurrence already increases strongly within 2–4 yr after conversion to organic farming, and only became slightly larger (reaching 80–90% of sites) with a longer time span since conversion. The majority of roots yielded good-quality sequences through direct Sanger sequencing suggesting low diversity of Sebacinales, even though our templates were pools of roots from six individual plants. This is in agreement with results obtained by Selosse et al. (2009) and Garnica et al. (2013) who could directly sequence PCR product from various plant species.
The mechanism responsible for this increased prevalence of Sebacinales in organically managed fields is uncertain; in organically managed fields no pesticides or mineral fertilizers are being used, but replaced with organic fertilizers, and it is thus possible that Sebacinales are sensitive to any of these factors. Another important difference is that the abundance of weeds was much higher in organically managed fields compared to the conventionally managed fields (Honegger et al., 2014). Studies of natural ecosystems including temperate grasslands (Wehner et al., 2014), arctic vegetation (Blaalid et al., 2014), and forest soil (Buée et al., 2009) have reported average Sebacinales read numbers to range between 1.7% and 11.3% of all fungi. Even though PCR based sequencing can only give a very rough estimate of relative abundances, the observation that in the current study read numbers were much lower (and never exceeded 1%) may indicate that Sebacinales are sensitive to agriculture in general. Given that this sensitivity appears to extend from organic to more intensive agricultural practices means that Sebacinales may be useful as bio-indicators (e.g. to detect the use of pesticides, mineral fertilizers, or presence of weeds).
The functional consequence of the finding that Sebacinales are possibly absent in intensively used agricultural sites is not certain; Sebacinales show a surprisingly wide spectrum of mycorrhizal types (ectomycorrhizas, orchid mycorrhizas, ericoid mycorrhizas, and others) but are also found as apparently symptomless endophytes in vascular plants of nearly all assessed plant families (Selosse et al., 2009; Oberwinkler et al., 2013). Members have been shown to have positive growth effects on cereals like maize and barley through a variety of mechanisms (Waller et al., 2005, 2008; Yadav et al., 2010; Oberwinkler et al., 2013). So far however, this work has mainly focused on the only two cultured endophytic Sebacinales, Piriformospora indica and Sebacina vermifera (Waller et al., 2008; Yadav et al., 2010). Waller et al. (2005) revealed that P. indica has biocontrol activities as its presence strongly suppressed the negative effects of the root pathogen Fusarium culmorum on barley. Whether similar functions are performed in the sites we sampled remains to be studied, as they were found at low relative abundance and represent thus far uncultured strains of Sebacinales.
More research is now needed to elucidate which factors control Sebacinales in roots of crop plants, of which organic farming practices and plant diversity are potentially fruitful candidates. In this respect it is especially important to test whether the Sebacinales present at low abundance in wheat and other crops have a biological function (e.g. acquire nutrients for the plant, provide protection from disease and stress), and whether sampling at multiple time-points throughout the season might change abundance and occurrence estimates. Our results are also interesting in the context of the recent finding that farming has potentially led to a loss of groups of microbiota which convey functions present in natural systems, from which they have developed (Fierer et al., 2013). Our results suggest that Sebacinales may be another compelling example of such losses.
The authors thank Hansruedi Oberholzer, Caroline Scherrer, Urs Zihlmann and Fredi Strasser for discussion and practical support, and Adrian Honegger for the plant species inventory. The authors want to thank three anonymous reviewers for their supportive comments and valuable improvements. E.V. acknowledges support from Freie Universität Berlin, D.H., R.W. and M.G.A.vdH. were funded by the Institute of Sustainability Sciences, Agroscope, and M.H. was supported by grant 137136 and 143097 from the Swiss National Science Foundation.